Pure mode II fracture characterization of composite bonded joints

M.F.S.F. de Moura
a,
*
, R.D.S.G. Campilho
a
, J.P.M. Gonalves
b
a
Departamento de Engenharia Mecnica e Gesto Industrial, Faculdade de Engenharia da Universidade do Porto, Rua Dr. Roberto Frias, s/n, 4200-465 Porto, Portugal
b
Mathematical Sciences Department, IBM T. J. Watson Research Center, 1101 Kitchawan Road, Yorktown Heights, NY 10598, USA
a r t i c l e i n f o
Keywords:
Adhesive joints
Fracture toughness
Finite element analysis
Cohesive damage models
a b s t r a c t
A new data reduction scheme is proposed for measuring the critical fracture energy of adhesive joints
under pure mode II loading using the End Notched Flexure test. The method is based on the crack equiv-
alent concept and does not require crack length monitoring during propagation, which is very difcult to
perform accurately in these tests. The proposed methodology also accounts for the energy dissipated at
the Fracture Process Zone which is not negligible when ductile adhesives are used. Experimental tests
and numerical analyses using a trapezoidal cohesive mixed-mode damage model demonstrated the good
performance of the new method, namely when compared to classical data reduction schemes. An inverse
method was used to determine the cohesive properties, tting the numerical and experimental loaddis-
placement curves. Excellent agreement between the numerical and experimental R-curves was achieved
demonstrating the effectiveness of the proposed method.
1. Introduction
Adhesively bonded joints present several advantages relative to
alternative joining methods. They present fewer sources of stress
concentrations, behave well under fatigue loads, and allow joining
different materials. However, their use on structural applications is
still limited. Most of the design approaches are based on strength
of materials concepts which are clearly inadequate when singular-
ities are present. In fact, stresses at the singularity points are mark-
edly mesh dependent when a nite element analysis is used. The
alternative is the use of fracture mechanics concepts where the
energetic based approaches acquire special relevancy relative to
the approaches based on stress intensity factors. Energetic analyses
provide a suitable measurement of the critical fracture energy
using non-local parameters such as applied load and displacement.
The fracture mechanics approaches rely on the denition of an ini-
tial aw or crack length. However, in many structural applications
the locus of damage initiation is not obvious. On the other hand,
stress-based methods behave well at predicting damage initiation,
and fracture mechanics behaves well in modelling damage propa-
gation. In order to overcome the drawbacks of each method and
exploit the usefulness of the described advantages, cohesive dam-
age models become suitable options (Blackman et al., 2003;
Andersson and Stigh, 2004; Lefer et al., 2007; de Moura, 2006).
These methodologies combine aspects of stress based analysis to
model damage initiation and fracture mechanics to deal with dam-
age propagation. Thus, it is not necessary to take into consideration
an initial defect and mesh dependency problems are overcome.
Nevertheless, cohesive models depend on accurate measurement
of the critical fracture energies to be used on chosen energetic
criteria.
Fracture characterization of bonded joints under pure mode I
has been extensively studied by several authors (Andersson and
Stigh, 2004; Bader et al., 2000; Ducept et al., 2000; Nairn, 2000).
However, mode II is still not well addressed owing to some partic-
ular aspects inherent to the most popular tests: the End Notched
Flexure (ENF), the End Loaded Split (ELS) and the Four-Point End
Notched Flexure (4ENF). The ELS test involves a clamp which is a
source of variability and increases the complexity of data reduction
(Blackman et al., 2005). On the other hand, the 4ENF test requires a
complex setup and presents some problems related to large fric-
tion effects (Schuecker and Davidson, 2000). As a consequence,
the ENF appears to be the most suitable test for mode II fracture
characterization of bonded joints (Lefer et al., 2007). However,
problems related to unstable crack growth and to crack monitoring
during propagation have not been adequately solved yet. In fact, in
the mode II fracture characterization tests, the crack tends to close
due to the applied load, which hinders a clear visualization of its
tip. In addition, the classical data reduction schemes, based on
beam theory analysis and compliance calibration, require crack
monitoring during propagation. On the other hand, a quite exten-
sive Fracture Process Zone (FPZ) ahead of crack tip exists under
mode II loading for ductile adhesives. This non-negligible FPZ
affects the measured fracture energy. Consequently, its inuence
should be taken into account, which does not occur when a real
crack length is used in the selected data reduction scheme. To over-
come these limitations, a new data reduction scheme based on the
* Corresponding author.
E-mail address: mfmoura@fe.up.pt (M.F.S.F. de Moura).
crack equivalent concept and depending only on the specimen
compliance is presented in this work. Similar approaches were pro-
posed by Tamuzs et al. (2003) and Biel and Stigh (2008) for the
Double Cantilever Beam (DCB) specimen. The main objective of
the proposed methodology is to increase the accuracy of the criti-
cal fracture energy measurements resulting from experimental
mode II tests. The method was applied to experimental tests and
validated by a numerical approach that uses a trapezoidal cohesive
mixed-mode damage model to account for the adhesive ductility.
The cohesive parameters dening the constitutive trapezoidal
law were obtained using an inverse method, tting the loaddis-
placement curves. Classical data reduction schemes based on spec-
imen compliance calibration and corrected beam theory were also
used and were compared with the proposed method.
2. Experimental work
Fig. 1 shows the geometry and dimensions of the ENF speci-
mens. In order to provide crack growth stability (Carlsson et al.,
1986), the initial crack length was considered to be equal to 70%
of the half-length of the specimen. Unidirectional 0 lay-ups of car-
bon/epoxy prepreg (TEXIPREG HS 160 RM) with 0.15 mm ply
thickness were used as adherends, whose mechanical properties
are presented in Table 1 (Campilho et al., 2005). The laminates
were manufactured by the hand lay-up technique and cured in a
hot-plates press during 1 h at 130 C and 4 bar pressure. The duc-
tile epoxy adhesive Araldite

2015, whose elastic properties were

measured experimentally in bulk tests (E = 1850 MPa, m = 0.3),
was used. The bonding surfaces were roughened with sandpaper
and cleaned with acetone prior to bonding, in order to avoid un-
wanted adhesive failures. Assembly was achieved by holding with
contact pressure and curing at room temperature. Five specimens
were tested using an INSTRON testing machine at room tempera-
ture under displacement control (2 mm/min). Fig. 2a shows the
experimental setup. The loaddisplacement (Pd) curve was regis-
tered during the test. In mode II fracture characterization tests,
crack length monitoring during propagation is very difcult to per-
form, as the crack grows without a clear opening (Blackman et al.,
2005; de Moura et al., 2006). Nevertheless, the crack length was
monitored by bonding to the specimens edge a strip of paper with
the graduations printed on it and by taking photos during the tests
with 5 s intervals using a 10 mega pixel digital camera (Fig. 2b). For
a better crack tip visualization a white correction uid layer was
used. These procedures help minimize the reading errors made
when measuring crack length by visual inspection during the
course of the test. The experimental values of Pda as a function
of time were obtained. The time of each Pd data point was calcu-
lated from the applied displacement and the chosen loading rate.
The time for each value of a is the one at which the corresponding
photo was taken.
3. Data reduction schemes
3.1. Classical methods
The classical data reduction schemes to obtain the critical frac-
ture energy in pure mode II (J
IIc
) are usually based on compliance
calibration or beam theory. The Compliance Calibration Method
(CCM) is based on the IrwinKies equation (Kanninen and Popelar,
1985)
J
IIc

P
2
2B
dC
da
1
where B is the width and C = d/P the compliance of the specimen.
Cubic polynomials (C = C
1
a
3
+ C
0
) were used to t the C = f(a) curves,
leading to
J
IIc

3P
2
C
1
a
2
2B
2
In the case of the ENF test, the Corrected Beam Theory (CBT) pro-
posed by Wang and Williams (1992) leads to
J
IIc

9a 0:42D
I

2
P
2
16B
2
h
3
E
1
3
where E
1
is the axial modulus, h is the half height of the specimen
and D
I
a crack length correction to account for shear deformation
D
I
h

E
1
11G
13
3 2
C
1 C
_ _
2
_ _

_
4
Fig. 1. Schematic representation of the ENF test (dimensions in mm).
Table 1
Carbonepoxy elastic properties.
E
1
= 109 GPa m
12
= 0.342 G
12
= 4315 MPa
E
2
= 8819 MPa m
13
= 0.342 G
13
= 4315 MPa
E
3
= 8819 MPa m
23
= 0.380 G
23
= 3200 MPa
5 mm crack tip
a b
Fig. 2. Experimental setup (a) and crack length measurement during propagation (b).
with
C 1:18

E
1
E
3
p
G
13
5
where E
3
and G
13
are the transverse and shear moduli,
respectively.
3.2. Compliance Based Beam Method
The methods described in the previous section require accurate
crack length measurements during propagation, which are not
easy to obtain. As a result, important errors can occur during frac-
ture characterization of bonded joints under pure mode II loading.
On the other hand, modern adhesives usually present a signicant
ductile behaviour. In these cases, a large FPZ develops ahead of the
crack tip, which is responsible for a non-negligible amount of en-
ergy dissipation. Consequently, the used data reduction schemes
should take this issue into account. The CBT depends on the real
crack length measured experimentally, thus not including the en-
ergy dissipated in the FPZ. The CCM is based on specimen compli-
ance calibration, which is affected by the energy being dissipated
in the FPZ. However, in view of the used cubic polynomial ap-
proach, the fracture energy equation also depends on the real crack
length (Eq. (2)).
In order to overcome these limitations, a data reduction scheme
based on the crack equivalent concept is proposed. It is named
Compliance Based Beam Method (CBBM) and depends only on
the specimens compliance during the test, which, using the beam
theory, can be written as
C
3a
3
2L
3
8E
1
Bh
3

3L
10G
13
Bh
6
Since the exural modulus of the specimen plays a fundamental
role on the Pd relationship, it can be calculated from Eq. (6) using
the initial compliance C
0
and the initial crack length a
0
E
f

3a
3
0
2L
3
8Bh
3
C
0corr
7
where C
0corr
is given by
C
0corr
C
0

3L
10G
13
Bh
8
This procedure takes into account the variability of the material
properties between different specimens and several effects that
are not included in beam theory, e.g., stress concentrations near
the crack tip, contact between the specimen arms at the pre-crack
region and root displacement and rotation effects. In fact, these
phenomena affect the specimen behaviour and consequently the
Pd curve, even in the elastic regime. Using this methodology, their
inuence is accounted for through the calculated exural modulus.
The effect of the FPZ can be included considering the com-
pliance and the equivalent crack concept during propagation.
Consequently, during crack propagation, a correction of the real
crack length is considered in the equation of compliance (6) to
account for the FPZ inuence. Substituting E
f
(Eq. (7)) and a
e
in
the place of E
1
and a, respectively, in Eq. (6) it can be written
a
e
a Da
FPZ

C
corr
C
0corr
a
3
0

2
3
C
corr
C
0corr
1
_ _
L
3
_ _
1=3
9
where C
corr
is given by Eq. (8) using C instead of C
0
. J
IIc
can now be
obtained using the IrwinKies relation (Eq. (1))
J
IIc

9P
2
a
2
e
16B
2
E
f
h
3
10
Using the methodology presented above, the critical fracture en-
ergy J
IIc
is obtained just from the Pd curve. Note that the modulus
of the specimen is not an inputted property but a computed one
(Eq. (7)) and is a function of the initial compliance and G
13
, which
is the only material property needed in this approach. Previous
studies (de Moura et al., 2006) showed that G
13
has much less
inuence than the longitudinal modulus, which means that a typ-
ical value can be used. Note also that in the above approach it is
not necessary to measure the crack length during propagation be-
cause the calculated equivalent crack length is used instead of the
real one. Another advantage is related to the fact that a
e
includes
the effect of the FPZ, which is not taken into account when the real
crack length is considered. Moreover, a complete R-curve can be
obtained using the proposed methodology. Lefer et al. (2007) also
proposed a methodology that does not require crack length mon-
itoring. However, shear deformation measurements at the crack
tip are required which are not easy to perform when thin adher-
ends are used.
4. Experimental results
The experimental Pda parameters were used to obtain the
critical fracture energy in pure mode II as a function of the crack
length using the CCM and the CBT. The crack length correction
D
1
used in the CBT was calculated individually for all specimens,
taking into account the Youngs modulus variation between speci-
mens. An average value of 4.706 mm was obtained. The complete
R-curve was determined using the CBBM from the respective
experimental Pd curves. The exural modulus of all tested speci-
mens, necessary for the CBBM, was calculated using Eq. (7) and an
average value of 107.5 GPa was obtained. This value is not very dif-
ferent than the nominal Youngs modulus presented in Table 1.
A comparison between the three methods is presented in Fig. 3
for one of the specimens tested. In order to plot the CBBM R-curve
as a function of a, it was necessary to establish a correspondence
between a
e
and a using the applied displacement. The global re-
sults of all specimens are summarized in Table 2. Fig. 3 shows,
for the three data reduction schemes, a short plateau region fol-
lowed by an increasing trend of J
II
= f(a), especially when the crack
approaches the loading cylinder. This effect is explained by the
large FPZ of this particular adhesive, due to its high fracture
toughness in pure mode II. In fact, the quite large FPZ quickly
attains the central loaded region where the local effects of com-
pressive stresses hinder self-similar crack propagation for higher
values of crack length. The CCM and the CBBM present a similar
plateau. The CBT underestimates the critical fracture energy (de
Moura, 2006). In analysing the average results of all specimens
(Table 2), J
IIc
is slightly higher for the CBBM and smaller for the
CBT.
0
1
2
3
4
5
6
65 75 85 95
a [mm]
J
I
I

[
N
/
m
m
]
CCM
CBT
CBBM
Fig. 3. Experimental R-curves obtained by the CCM, CBT and CBBM.
5. Numerical analysis
With the objective of numerically simulating the behaviour of
the adhesive layer in pure mode II, a trapezoidal cohesive damage
model was developed and implemented within interface nite ele-
ments in the ABAQUS

software. The elements have zero thickness

and intend to replace the solid nite elements traditionally used to
represent the adhesive layer. The complete cohesive law in pure
mode II was determined using an inverse method, tting the
numerical and experimental Pd curves. This law was then used
to assess the adequacy of the three data reduction schemes utilized
to obtain J
IIc
. This assessment is based on the comparison between
the values of J
IIc
inputted in the numerical analyses and the values
provided by the three methods applied to the Pda results
obtained from those analyses.
5.1. Trapezoidal cohesive damage model
A mixed-mode (I + II) cohesive damage model implemented
within interface nite elements was developed to simulate damage
onset and growth. The adhesive layer is simulated by these
elements, which have zero thickness. To simulate the behaviour
of ductile adhesives, a trapezoidal softening law relating stresses
(r) and relative displacements (d
r
) between homologous points
of the interface elements was employed (Fig. 4). These types of
laws accurately reproduced the behaviour of thin ductile adhesive
layers in mode I (Andersson and Stigh, 2004) and mode II (Lefer
et al., 2007). The constitutive relationship before damage onset is
r Ed
r
11
where E is a stiffness diagonal matrix containing the stiffness
parameters e
i
(i = I, II), dened as the ratio between the elastic mod-
ulus of the material in tension or shear (E or G, respectively) and the
adhesive thickness t. Considering the pure-mode model, after d
1,i
(the rst inexion point, which leads to the plateau region of the
trapezoidal law) the material softens progressively. The softening
relationship can be written as
r I DEd
r
12
where I is the identity matrix and D is a diagonal matrix containing,
on the position corresponding to mode i (i = I, II), the damage
parameter. In general, bonded joints are under mixed-mode load-
ing. A formulation for interface nite elements should therefore
include a mixed-mode damage model (Fig. 4), which can also be ap-
plied under pure mode loading. In fact, the pure mode II loading
characteristic of the ENF test is a particular case of a general
mixed-mode loading.
Damage onset is predicted using a quadratic stress criterion
r
I
r
u;I
_ _
2

r
II
r
u;II
_ _
2
1 if r
I
> 0
r
II
r
u;II
if r
I
6 0
13
where r
i
(i = I, II) represent the stresses in each mode. It is assumed
that normal compressive stresses do not induce damage. Consider-
ing Eq. (11), the rst part of Eq. (13) can be rewritten as a function
of the relative displacements
d
1m;I
d
1;I
_ _
2

d
1m;II
d
1;II
_ _
2
1 14
where d
1m,i
(i = I, II) are the relative displacements in each mode
corresponding to damage initiation. Stress softening onset (d
2,i
)
was predicted using a quadratic relative displacements criterion
similar to (14), leading to
d
2m;I
d
2;I
_ _
2

d
2m;II
d
2;II
_ _
2
1 15
where d
2m,i
(i = I, II) are the relative displacements in each mode
corresponding to stress softening onset. Crack growth was simu-
lated by the linear fracture energetic criterion
J
I
J
Ic

J
II
J
IIc
1 16
When Eq. (16) is satised damage growth occurs and stresses are
released, with the exception of normal compressive ones. Using
the proposed criteria (Eqs. (14)(16)), it is possible to dene the
equivalent mixed-mode displacements (d
1m
, d
2m
and d
um
) and to
establish the damage parameter in the plateau region
d
m
1
d
1m
d
m
17
and in the stress softening part of the cohesive law
d
m
1
d
1m
d
um
d
m

d
m
d
um
d
2m

18
The damage parameter is introduced in Eq. (12), thus simulating
damage propagation. A detailed description of the proposed model
is presented in the work of Campilho et al. (2008).
5.2. Evaluation of the cohesive parameters
The value of G was obtained experimentally from E and m
which were determined from adhesive bulk tests and was used
to dene the stiffness parameters e
II
. The values of J
IIc
for the ve
specimens tested were obtained from the respective experimen-
tal loaddisplacement curves using the CBBM. The fracture
energy, which corresponds to the plateau value of the R-curves
(Table 2), was an inputted parameter in the numerical approach.
The remaining cohesive parameters (r
u,II
and d
2,II
) were deter-
mined by an inverse method, tting the numerical and experi-
mental Pd curves of each specimen. Fig. 5 shows the numerical
Table 2
Fracture energies in pure mode II obtained by the CCM, CBT and CBBM (N/mm).
CCM CBT CBBM
1 4.71 4.7 4.94
2 4.87 4.87 5.13
3 4.81 4.09 4.82
4 4.15 4.27 4.28
5 4.44 4.09 4.32
Average J
IIc
4.60 4.40 4.70
Standard deviation 0.27 0.32 0.34

u,i

um,i

i

1m,i

1,i

um,i

u,i

i
Pure-mode
model
Mixed-
mode
J
ic
(i = I, II)
J
i

2,i

2m,i
(i = I, II)
Fig. 4. The trapezoidal softening law for pure-mode and mixed-mode.
and experimental Pd curves for one tested specimen after the t-
ting procedure.
Fig. 6 shows the deformed shape of the ENF specimen during
crack propagation, and the respective boundary and loading condi-
tions. The specimen arms were modelled with plane strain eight-
node quadrilateral solid nite elements (CPE8 from the ABAQUS

library). The adhesive layer was simulated with six-node interface

nite elements compatible with the ABAQUS

elements, including
the trapezoidal mixed-mode cohesive damage model. Each speci-
men arm was modelled by eight solid nite elements through-
thickness. A more rened mesh was considered at the propagation
region and near the cylinders. Boundary conditions included xing
the supporting cylinders in the directions x and y and restraining
the loading cylinder in the direction x. The lowest node at the spec-
imen mid-section was restrained in the direction x.
Fig. 7 shows the average values of J
IIc
, r
u,II
and d
2,II
and the trap-
ezoidal cohesive laws range obtained by tting the ve experimen-
tal Pd curves. All of these parameters inuence the numerical Pd
curves prole. J
IIc
, which is the inputted value in the numerical
simulations, mainly inuences the peak load value. Higher local
strengths (r
u,II
) increase the peak load, and the specimen stiffness
up to the peak load, leading to a more abrupt post-peak load reduc-
tion. Finally, d
2,II
plays an important role on the roundness form at
the peak value of the Pd curve.
6. Comparison between the numerical and experimental results
A study was performed numerically to assess the adequacy of
the three data reduction schemes evaluated to measure J
IIc
accu-
rately. The objective is to verify how the used methods reproduce
the inputted J
IIc
. To accomplish this study, the numerical Pda
parameters were collected during crack propagation and were
used to obtain the critical fracture energies. Fig. 8 shows the results
for one case. Accurate results were obtained with the CCM and
especially the CBBM. The CBT underestimated the inputted J
IIc
,
which is explained by the non-negligible amount of energy being
dissipated in the FPZ that is not accounted for in this method.
Moreover, it should be emphasized that the CCM also requires
crack length measurements during the propagation stage, which,
even using an optical method, is prone to introduce additional
errors. On the other hand, using the CBBM, a complete R-curve is
obtained only from Pd data. Fig. 9 compares the numerical and
experimental R-curves using the CBBMfor the same specimen used
in Fig. 5, after the tting procedure. In this case J
II
is plotted as a
function of equivalent crack instead of real crack length, since
the objective of this method is to avoid crack length measure-
ments. As a consequence the R-curve presents a different prole.
The initial part of the curve (before the plateau) does not corre-
spond to crack propagation, but to FPZ development. In fact, as a
e
0
200
400
600
800
0 2 4 6 8 10 12 14
[mm]
P

[
N
]
Experimental
Numerical
Fig. 5. Comparison between the numerical and experimental Pd curves.
Fig. 6. ENF specimen during crack propagation and BCs.
0
10
20
30
40
0
[mm]

[
M
P
a
]
J
IIc avg
=4.70 N/mm

2,II avg
=0.171 mm

u,II avg
=22.8 MPa

u,II avg
=
0.248 mm
0.05 0.1 0.15 0.25 0.2 0.3
Fig. 7. Trapezoidal cohesive laws range obtained and cohesive parameters average
values.
0
1
2
3
4
5
6
65
a [mm]
J
I
I

[
N
/
m
m
]
Inputted value
CCM
CBT
CBBM
75 85
Fig. 8. Numerical R-curves obtained by the CCM, CBT and CBBM.
includes the FPZ effects, its variation up to the plateau region is
only due to FPZ development. The crack growth only occurs at
the beginning of the plateau region. A summary of the J
IIc
values
predicted by the several data reduction schemes for the ve spec-
imens is presented in Table 3. Accurate predictions were obtained
with the CBBM (average error of 0.6%) and the CCM (average error
of 2.9%). However, a large discrepancy was obtained with the CBT,
which underestimated the inputted J
IIc
by 24.9%. This behaviour is
also very different from the one obtained experimentally where
the values provided by the three methods were closer. The expla-
nation for this difference is given in Fig. 10. As it can be seen, for
the same specimen compliance there is an almost constant dis-
crepancy between the experimental and numerical crack length.
This statement leads to the conclusion that some problems oc-
curred during the experimental identication of the crack tip.
Probably, the damage ahead of the crack tip induces material
deformation originating some local slip, which fractures the white
correction uid layer used to better follow the crack length. This
phenomenon yielded a measured crack longer than the real one
by a constant amount, during the process of self-similar propaga-
tion with a pronounced FPZ at the crack tip.
7. Conclusions
A methodology for fracture characterization of ductile adhesive
layers under pure mode II used in bonded joints was proposed. The
respective critical fracture energy (J
IIc
) was measured by ENF tests,
using a new data reduction scheme based on the crack equivalent
concept. The results were compared with two classical data reduc-
tion schemes, CCM and CBT. The main advantages of the proposed
method include the absence of the need to monitor crack length
during its growth and the inclusion of the energy dissipated at
the FPZ, which can be non-negligible when ductile adhesives are
used. In order to evaluate the adequacy of the several data reduc-
tion schemes to measure J
IIc
, a numerical study was carried out
including a developed trapezoidal mixed-mode cohesive damage
model to simulate the behaviour of ductile adhesives. An inverse
method was used to dene the remaining cohesive parameters of
the trapezoidal law, tting the numerical and experimental Pd
curves.
It was veried that the CBBM renders the most accurate results
and is a suitable method. Consequently, and due to its advantages,
it was considered by the authors to be the best choice to be used on
the fracture characterization of bonded joints.
Acknowledgments
The rst author thanks the Portuguese Foundation for Science
and Technology for supporting the work presented here, through
the research Project POCI/EME/56567/2004. The second author
thanks the Portuguese Foundation for Science and Technology for
supporting the work presented here through the individual Grant
SFRH/BD/30305/2006.
References
Andersson, T., Stigh, U., 2004. The stresselongation relation for an adhesive layer
loaded in peel using equilibrium of energetic forces. International Journal of
Solids and Structures 41, 413434.
Bader, M.G., Hamerton, I., Hay, J.N., Kemp, M., Winchester, S., 2000. Double
cantilever beam testing of repaired carbon bre composites. Composites A 31,
603608.
Biel, A., Stigh, U., 2008. Effects of constitutive parameters on the accuracy of
measured fracture energy using the DCB-specimen. Engineering Fracture
Mechanics 75, 29682983.
Blackman, B.R.K., Hadavinia, H., Kinloch, A.J., Williams, J.G., 2003. The use of a
cohesive zone model to study the fracture of bre composites and adhesively
bonded joints. International Journal of Fracture 119, 2546.
Blackman, B.R.K., Kinloch, A.J., Paraschi, M., 2005. The determination of the mode II
adhesive fracture resistance, G
IIc
, of structural adhesive joints: an effective crack
length approach. Engineering Fracture Mechanics 72, 877897.
Campilho, R.D.S.G., de Moura, M.F.S.F., Domingues, J.J.M.S., 2005. Modelling single
and double-lap repairs on composite materials. Composites Science and
Technology 65, 19481958.
Campilho, R.D.S.G., de Moura, M.F.S.F., Domingues, J.J.M.S., 2008. Using a cohesive
damage model to predict the tensile behaviour of CFRP single-strap repairs.
International Journal of Solids and Structures 45, 14971512.
Carlsson, L.A., Gillespie, J.W., Pipes, R.B., 1986. On the analysis and design of the end
notched exure (ENF) specimen for mode II testing. Journal of Composite
Materials 20, 594604.
de Moura, M.F.S.F., 2006. Numerical simulation of the ENF test for the mode-II
fracture characterization of bonded joints. Journal of Adhesion Science and
Technology 20, 3752.
de Moura, M.F.S.F., Silva, M.A.L., de Morais, A.B., Morais, J.J.L., 2006. Equivalent crack
based mode II fracture characterization of wood. Engineering Fracture
Mechanics 73, 978993.
Ducept, F., Davies, P., Gamby, D., 2000. Mixed mode failure criteria for a glass/epoxy
composite and an adhesively bonded composite/composite joint. International
Journal of Adhesion and Adhesives 20, 233244.
Kanninen, M.F., Popelar, C.H., 1985. Advanced Fracture Mechanics. Oxford
University Press, Oxford.
Lefer, K., Alfredsson, K.S., Stigh, U., 2007. Shear behaviour of adhesive layers.
International Journal of Solids and Structures 44, 530545.
0
1
2
3
4
5
6
70
a
e
[mm]
J
I
I
c

[
N
/
m
m
]
Experimental
Numerical
75 80 85 90
Fig. 9. Comparison between the numerical and experimental R-curves using the
CBBM.
Table 3
Inputted and predicted J
IIc
(N/mm) values using the three data reduction schemes.
Specimen Inputted CCM error (%) CBT error (%) CBBM error (%)
1 4.94 4.74 4.0 3.73 24.5 4.89 0.9
2 5.13 5.05 1.6 3.84 25.1 5.07 1.1
3 4.82 4.76 1.2 3.62 24.9 4.80 0.5
4 4.28 4.10 4.2 3.24 24.3 4.26 0.4
5 4.32 4.17 3.5 3.21 25.7 4.31 0.2
Average 4.56 2.9 3.53 24.9 4.67 0.6
60
65
70
75
80
85
90
95
0.015
C [mm/N]
a

[
m
m
]
Experimental
Numerical
0.017 0.019 0.021 0.023 0.025
Fig. 10. Numerical and experimental crack length as a function of the specimen
compliance.
Nairn, J.A., 2000. Energy release rate analysis for adhesive and laminate double
cantilever beam specimens emphasizing the effect of residual stresses.
International Journal of Adhesion and Adhesives 20, 5970.
Schuecker, C., Davidson, B.D., 2000. Effect of friction on the perceived mode II
delamination toughness from three- and four-point bend end notched exure
tests. ASTM STP 1383, 334344.
Tamuzs, V., Tarasovs, S., Vilks, U., 2003. Delamination properties of
translaminar-reinforced composites. Composites Science and Technology
63, 14231431.
Wang, Y., Williams, J.G., 1992. Corrections for mode II fracture toughness
specimens of composite materials. Composites Science and Technology 43,
251256.